"Arne Vajhøj" <> wrote in message
news:4b8afff4$0$272$...
> On 28-02-2010 15:19, DuncanIdaho wrote:
>> Well. thank's for all your comments. I've always used the development
>> kits available from the apparently now defunct http://java.sun.com which
>> now appears as an Oracle branded web site.
>
> Ofcourse.
>
> When Oracle buys SUN then they obviously rebrand all the SUN web sites,
>
> That is to be expected.
>
yep.
> > Anyway as for IBM, I remember using some
>> ghastly piece of IBM IDE nastiness whose name I have expunged from my
>> memory forever. It was an explosion of green and grey with millions of
>> windows and the most appalling spaghetti code nonsense you ever have
>> seen ...
>
> Visual Age for Java 3.5 and 4.0 is on the top-10-hate-list of
> many developers.
>
can't claim any personal experience with this one...
> > and if I ever had to 'migrate' to .NET I think I'd drown myself
>
> It is not that bad.
>
> You will find a lot of the stuff very familiar.
>
agreed.
actually, MS's tools and basic technology are fairly well polished.
Visual Studio works fairly well (although, for other reasons, it doesn't
entirely sway me from my primary use of Makefile's for building code...).
similarly, C# is a fairly capable language (and not that much different from
Java in many respects), and at well, C and C++ can also work on .NET
(although one does have to force the C into pretending to be C++, but this
is not too big of an issue...).
CIL (AKA: MSIL) is also a fairly capable bytecode, as it can also directly
handle C and C++, as well as C#, J#, VB, ...
also, C++ on .NET (C++/CLI) is also capable of fairly directly accessing
code and data in native land (no need for something like JNI or JNA).
technically, CIL is actually a bit more capable and powerful than JBC (JBC
is not particularly well suited to languages which are not Java...).
granted, some non-Java languages have used JBC, but generally not cleanly
(hacks are often needed), and a language like C or C++ can't be run directly
(it has to be "faked" in many ways). CIL is, however, directly capable of
representing C or C++ semantics (raw pointers, function pointers, ...) at
the IL level.
but, what of the cost?:
between them, CIL is very much more complex, and its binary-representation
(as a hacked-over PE/COFF image with something vaguely similar to a
relational-database thrown in the mix), is generally somewhat ugly and
difficult to work with (difficult to load, difficult to process, metadata
has an inconvinient layout and is not really easy to convert, ...).
similarly, there are not any particularly good alternative implementations
(many alternative implementations exist of Java, but very few of .NET, and
those which do exist are generally fairly poor...).
but, I have gone and tried to make a VM which can deal with both JBC and CIL
(as well, it has its own internal IL as well, of a vaguely similar style to
JBC and CIL, but this IL is not particularly well suited for anything not
C...).
in general, the VM tries to share most of the internal machinery (mostly by
me trying to keep everything modular and general-purpose), ...
recently, I had also been trying to write my own implementation of the Java
class library, so I can have one I can put under public domain, but then I
am faced with what would seem to be the apparent task of writing a mountain
of boilerplate (my experience thus far has been writing lots of code which
seems to do relatively little, I have little idea why so many classes and
methods are needed for all this stuff...).
granted, I may be missing something here, as I primarily code in languages
which are not Java...
or such...
> Arne
Similar Threads
